Reconstruction of motor control circuits in adult Drosophila using automated transmission electron microscopy

Abstract

To investigate circuit mechanisms underlying locomotor behavior, we used serial-section electron microscopy (EM) to acquire a synapse-resolution dataset containing the ventral nerve cord (VNC) of an adult female Drosophila melanogaster. To generate this dataset, we developed GridTape, a technology that combines automated serial-section collection with automated high-throughput transmission EM. Using this dataset, we studied neuronal networks that control leg and wing movements by reconstructing all 507 motor neurons that control the limbs. We show that a specific class of leg sensory neurons synapses directly onto motor neurons with the largest-caliber axons on both sides of the body, representing a unique pathway for fast limb control. We provide open access to the dataset and reconstructions registered to a standard atlas to permit matching of cells between EM and light microscopy data. We also provide GridTape instrumentation designs and software to make large-scale EM more accessible and affordable to the scientific community.

Keywords: campaniform sensilla; connectomics; limb control; motor neurons; sensory feedback; serial-section electron microscopy.

Figure 1.. A high-throughput serial-section transmission electron microscopy (TEM) pipeline built around GridTape.

(A) Regularly spaced holes and barcodes are laser-milled into a length of tape to produce GridTape, a substrate for collection of serial sections. (B) Schematic of stacked GridTape layers in cross-section. Tape thickness is exaggerated for clarity. (C-E) Schematics of sectioning (top) and imaging (bottom) for different serial-section EM approaches. Bottom schematics do not share the same scale. (C) Manual serial-section collection and TEM imaging. Samples are serially sectioned and manually picked up onto coated slot grids (3 mm outer diameter). (D) Automated tape-collecting ultra-microtome (ATUM) serial-section collection and SEM imaging. Sections are collected automatically onto tape (8 mm wide). Tape is then cut into strips and adhered to a wafer (bottom) for imaging. Bottom inset: zoomed-in view of a section on tape. (E) GridTape serial-section collection and TEM imaging. Samples are sectioned using a GridTape-compatible ATUM. Sections adhere to GridTape (8 mm wide) immediately after being cut and are targeted to land over film-coated holes in the tape. GridTape-collected sections are imaged using a reel-to-reel system. Bottom inset: zoomed-in view of a section on GridTape. (F) Schematic of the GridTape reel-to-reel stage. Reels of GridTape are inserted into the custom stage, which positions sections under the electron beam. Portions of the TEM microscope column in beige. Electron beam in light blue (not to scale). Scale bars, 10 mm (A), 10 cm (F). Scale box, 1 mm (C-E, top).

Figure 2.. An adult Drosophila ventral nerve cord (VNC) dataset.

(A) Schematic of the adult Drosophila central nervous system and leg. The synapse-resolution EM dataset presented here contains the VNC and its connection to the brain (dashed outline). (B) The VNC was cut into 4355 thin sections and collected onto GridTape. Each black rectangle indicates the imaged region for a single section relative to the slot (orange outline). Two sections collected off-slot are not shown. (C) Volumetric rendering of the VNC dataset. Light grey, outline of all imaged tissue. Dark grey, outline of VNC neuropil. (D) A single coronal section (left, section 1228) and sagittal reslice through the aligned image volume (right). Green and purple dashed lines in (C) and (D) indicate the slice locations. The imaged region spans from the subesophageal ganglion in the ventral brain, across the neck connective to the metathoracic neuromere and the metathoracic leg nerve. (E) Zoomed-in sagittal reslice of the region (cyan box) in (D). (F) Zoom-in of the region (pink box) in (D). (G) Zoom-in of the region (yellow box) in (F) showing synapses. Yellow arrowheads indicate presynaptic specializations known as T-bars. Scale bars, 500 μm (B), 50 μm (D), 10 μm (F), 1 μm (E, G).

Figure 3.. Reconstruction of motor and sensory neurons reveals precise functional domains in nerves.

(A) All 507 motor neurons (MNs) in the VNC’s thoracic segments were reconstructed from the EM dataset. Each MN projects its axon out one peripheral nerve, leaving the EM dataset, to innervate muscles. Spheres represent cell bodies. Unless otherwise noted, all renderings are viewed from the dorsal side of the VNC. (B) 655 reconstructed sensory axons. Reconstruction included some neurons from all limbs but focused primarily on the left T1 neuromere (asterisk). Same color code as (A). (C) Sections through the prothoracic (T1), mesothoracic (T2), and metathoracic (T3) leg nerves, which contain most of the sensory and motor axons connecting the VNC to the front, middle, and hind legs, respectively. Section locations indicated by dashed boxes in (B). The leg nerves have distinct domains containing the axons of MNs (cyan) and sensory neurons (magenta). The only intermingling between motor and sensory axons is a group of three sensory axons within the motor domain of the T2 leg nerve (magenta arrowhead). (D) Reconstructions transformed into a standard atlas coordinate space (Fig. S3). Renderings of EM reconstructions in subsequent figures were transformed into the atlas space. Scale bars, 100 μm (A-B), 10 μm (C) 50 μm (D).

Figure 4.. Identification of sensory neuron subtypes.

(A) Reconstruction of the main branches of sensory axons for the front left leg. The four main functional subtypes of sensory neurons (different colors) are identifiable from their projection patterns. Light grey, VNC. Darker grey, neuropil. ProAN, prothoracic accessory nerve; ProLN, prothoracic leg nerve; VProN, ventral prothoracic nerve; DProN, dorsal prothoracic nerve. (B) Organization of hair plate neuron projections. Hair plate axons enter the T1 neuromere through four different nerves (different colors) and branch to encircle the neuromere. (C) Femoral chordotonal organ (FeCO) neuron subtypes. Inset: Different subtypes, characterized previously with light microscopy (LM), encode different aspects of leg kinematics (adapted from Mamiya et al., 2018). (D-E) Comparison of EM reconstructions with LM reconstructions from genetic driver lines that specifically label FeCO neurons (Mamiya et al. 2018, Chen et al. in preparation). (i) Rendering of LM reconstruction. (ii) Ranked distribution of NBLAST similarity scores (worst to best, left to right) color coded by FeCO neuron subtype (as in C). (iii) Overlay of the LM reconstruction and the five most similar EM reconstructions. (iv) The five most similar EM reconstructions alone. (D) A club FeCO neuron (MCFO from R64C04-Gal4). (E) A claw FeCO neuron (MCFO from iav-Gal4). (F) A hook FeCO neuron (R70H02-AD, R32H08-DBD). Scale bars, 50 μm (A-F).

Figure 5.. Bilateral campaniform sensillum (bCS) neurons from both sides of the body directly connect to MNs near their spike-initiation zones.

(A) Single bCS axons from the front (i), middle (ii), and hind (iii) left legs. Asterisks denote where each axon enters the VNC. (B) Two neurons with the morphologies shown in (A) originate from each of the six legs. Dashed boxes indicate a ~25 μm-long tract where bCS axons originating from one leg travel alongside bCS axons originating from other legs. (C) Right mesothoracic (T2) leg nerve. bCS axons (yellow) have large-caliber axons compared to other leg sensory and motor neurons. (D) Cross-sectional areas of bCS axons and MN primary neurites for three different legs. (E) A split-Gal4 line labeling bCS neurons. Full expression pattern (left) and a single bCS axon labeled using MultiColor FlpOut (Nern et al., 2015). (F) Lateral view of the location indicated by the arrowhead in (B). A, anterior; P, posterior; V, ventral; D, dorsal. In the boxed region, bCS axons originating from left T1 (dark red), right T1 (light red), and left and right T2 (not shown) converge, traveling directly alongside primary neurites of ProLN MNs (grey; same neurons as Fig. S6C). bCS output synapses denoted by cyan spheres. (G) Synapse from a right T2 bCS axon (yellow) onto two left T1 MNs (cyan). Arrowhead indicates presynaptic T-bar structure. All 12 bCS neurons synapse onto MNs in each neuromere to which they project. (H-I) Analysis of all synaptic connections made by left and right T1 bCS axons along the ~25 μm stretch indicated in (F). (H) Connections from left T1 versus right T1 bCS axons onto left ProLN MNs. The two left bCS axons and two right bCS axons largely target the same MNs. (I) Distribution of distances from each bCS synapse to each postsynaptic MN’s primary neurite (red, n=264 postsynaptic sites) compared to synapses randomly distributed across MN dendrites (cyan). Scale bars, 50 μm (A-B, E-F), 10 μm (C), 500 nm (G).

Figure 6.. MN bundles, symmetry, uniqueness, and bCS connectivity.

(A) Reconstruction of cell bodies and primary neurites of the 24 ProAN MNs (12 per side). Primary neurites travel through the neuromere in five distinct and highly symmetric bundles (numbered A1 through A5, colored in shades of purple). See also Fig. S6. (B) Quantitative analysis of bundles of MN primary neurites. ProAN MNs on each side of the VNC were clustered by the similarity in primary neurite positions (STAR Methods). Top, dendrogram from hierarchical clustering. Members of each bundle cluster together. Bottom, matrix of NBLAST similarity scores. (C) Branching patterns of all 139 MNs arborizing in the T1 neuromeres were reconstructed and transformed into the atlas coordinate system (Fig. S3). (D) Identification of left–right homologous pairs of front leg MNs. Of the 69 left and 70 right T1 MNs, expert annotators identified 61 symmetric left–right pairs. A global pairwise assignment of NBLAST similarity scores agreed on 92% (56 of 61) of identified pairs. Black asterisks, agreements. Red asterisks, disagreements. (E) Relationship between four anatomical properties of leg MNs: proximity to bCS axons (x-axis), primary neurite cross-sectional area (y-axis), primary neurite bundle (marker color), and number of synapses received from bCS neurons (marker type and size). MNs closer to bCS axons have larger-caliber primary neurites. Only MNs in the L1 bundle received any synapses from bCS neurons. Within the L1 bundle, those receiving the most synapses have large-caliber primary neurites and are closest to bCS axons. Dashed circle indicates the MN whose morphology is most similar to a functionally characterized fast flexor MN (Fig. 7A). Scale bars, 50 μm (A, C).

Figure 7.. A fast tibia flexor motor neuron is a major synaptic target of bCS neurons.

(A-C) MNs reconstructed from LM matched to the most similar neurons reconstructed from EM. (i) Rendering of LM reconstruction. (ii) Ranked distribution of NBLAST similarity scores (worst to best, left to right) color coded by MN bundle (key, Aii top). (iii) Zoom-in on the 8 highest similarity scores. (iv) Overlay of the LM reconstruction and the most similar EM reconstruction. (v) The most similar EM reconstruction. (vi) The second-most similar EM reconstruction. (A) A fast tibia flexor MN (81A07-Gal4). The two most similar EM reconstructions both receive strong synaptic input from the two left and two right T1 bCS neurons. (B) A slow tibia flexor MN (35C09-Gal4). The two most similar EM reconstructions receive minimal synaptic input from T1 bCS neurons. (C) A MN innervating the tibia long tendon muscle, which controls movements of the tarsus (21G01-LexA). The two most similar EM reconstructions receive no synaptic input from T1 bCS neurons. Scale bars, 50 μm (A-C).